
# imports
import numpy as np
from PIL import Image
import os
import glob
import matplotlib.pyplot as plt
import warnings

# Ignore all warnings
warnings.filterwarnings('ignore')
%matplotlib inline


# Read images and read all sizes?
# plot image sizes
image_path_list = []
hw_list = []
for image_path in glob.glob('./dataset/*/*/*.*'):
    image_path_list.append(image_path)
    with Image.open(image_path) as img:
        width, height = img.size
        hw_list.append([width, height])

Total_number_images = len(image_path_list)
print(f'This dataset has {len(image_path_list)} images.')

This dataset has 377 images.


# Randomize the samples
fig = plt.gcf()
fig.set_size_inches(20, 20)
for i, img_path in enumerate(image_path_list[:36]):
    sp = plt.subplot(6, 6, i + 1)
    sp.axis('Off')
    img = Image.open(img_path)
    plt.imshow(img)


# Extract height and width from the list of image dimensions
heights = [dim[0] for dim in hw_list]
widths = [dim[1] for dim in hw_list]

# Create scatter plot
plt.figure(figsize=(8, 6))
plt.scatter(widths, heights, color='blue', alpha=0.5)
plt.title('Image Dimensions Scatter Plot')
plt.xlabel('Width')
plt.ylabel('Height')
plt.grid(True)
plt.show()


# Plotting distribution of widths and heights
fig, ax = plt.subplots(1, 2, figsize=(15, 5))

# Plotting widths
ax[0].hist(widths, bins=30, color='skyblue', edgecolor='black')
ax[0].set_title('Distribution of Image Widths')
ax[0].set_xlabel('Width')
ax[0].set_ylabel('Number of Images')

# Plotting heights
ax[1].hist(heights, bins=30, color='salmon', edgecolor='black')
ax[1].set_title('Distribution of Image Heights')
ax[1].set_xlabel('Height')
ax[1].set_ylabel('Number of Images')

plt.tight_layout()
plt.show()


import pytesseract

image_dir = './dataset'

# Directory where you want to save the .txt files
output_dir = './dataset_text'

# Loop through each image file in the directory
for image_path in image_path_list:
        image = Image.open(image_path)
        
        # Extract text from the image
        extracted_text = pytesseract.image_to_string(image)

        # Close the image
        image.close()
        
        # Save the extracted text
        text_file_path = os.path.join(output_dir, os.path.basename(image_path) + '.txt')

        with open(text_file_path, 'w') as text_file:
            text_file.write(extracted_text)

        print(f"Text extracted from {image_path} and saved to {text_file_path}")


# examples of the extracted text
sample_text_path = 'dataset_text/350.jpg.txt'

with open(sample_text_path, 'r') as file:
        # Read the contents into a string
        file_contents = file.read()

print(file_contents.strip())

epick up our tele:
‘phone and dial an aie
line reservation number.

Bat instead of geting
recording asking you 9
‘wait, you get one offering
tohelp.

‘Youthen entera

MADE THE TELEPHONE |
In fact, right now, the

MAKING THE COMPUTER ©
power of a computer with

the sound of your voice.

puter?

      

seven-digit identification | and computers. reservations, administra- Each is becoming the
conde (suchas yourphone |" Werecreating data | on, et, other And becoming. in
number networks that do even magineifallthese | the process, more usetal
Youenterthe day | moreforinformation than | workedasonenetworkin | that ever betore
and month you want tofly. | our telephone network ‘concert with the data
‘You press buttons does for your voice. networks at car rental and
toindicateyourplace of | “These arenetworks | hotel chains.
departure. that not only move in “Travel wouldbe easier
Others toindicate your | formation instantly, but than you ever thought
destination, Thenumber | alsointerpetit,rearange | possible.
of persons travelling. The | it, and apply itin the most Infact, data network- ——
time ofthe fight. ‘seful ay will make all kinds ==
And you hang up with All automatically. ‘personal services more = SS
‘your reservation made. ‘Their potential value Senet enki >
Didyouuseaphone | to airlines extends realing, health car, ——
ora computer beyond ticketing. Most | home buying, investing,
Ifthe answer isnt | airlines have several credit approval, even Atel
clear, it's because at AT&T | separate voice and data | home entertainment.

weremening telephones | networks—forschedules, | Computers willfinally | Ghexaiglaticerrd@om


# Show image examples with text:

import matplotlib.pyplot as plt
from PIL import Image

# TODO clean extracted text

# Image examples are manually chosen
image_expl = ['dataset/subfolder-0/0/230.jpg', 'dataset/subfolder-0/0/50.jpg', 'dataset/subfolder-0/0/240.jpg']
corresponding_text_expl = ['dataset_text/230.jpg.txt', 'dataset_text/50.jpg.txt', 'dataset_text/240.jpg.txt']

# Open the file
for i in range(len(corresponding_text_expl)):
    with open(corresponding_text_expl[i], 'r') as file:
        # Read the contents into a string
        file_contents = file.read()

    image_path = image_expl[i]
    image = Image.open(image_path)

    fig, ax = plt.subplots()
    ax.imshow(image)
    ax.axis('off')

    # Display description next to the image
    ax.text(1.05, 0.5, file_contents, verticalalignment='center', transform=ax.transAxes, fontsize=12)

    plt.show()

Comprehensive Overview of the Advertisements Dataset¶

Overview¶

Dataset Samples¶

Insightful Visualization: Exploring Dataset Variation Through Image Dimensions¶

Extracting Textual Content with OCR for Dataset Understanding¶

Show image samples next to extracted text¶

Note¶